Dynamics of poly(methyl methacrylate) chains adsorbed on aluminum

polymer chains in nonequilibrium states is a result of the topography of the ..... 5, 1993. Dynamics of PMMA Chains Adsorbed on A1 Surfaces. 1123 trea...
0 downloads 0 Views 2MB Size
Macromolecules 1993,26, 1120-1136

1120

Dynamics of Poly(methy1 methacrylate) Chains Adsorbed on Aluminum Surfaces J. Scott Shaffer and Arup K.Chakraborty' Department of Chemical Engineering, University of California at Berkeley, and Center for Advanced Materials, Lawrence Berkeley Laboratory, Berkeley, California 94720-9989 Received September 8, 1992; Revised Manuscript Received November 16, 1992

ABSTRACT We present a stochasticmodel for the dynamics of poly(methy1methacrylate) (PMMA)chains adsorbed on an aluminum surface. The model is able to investigate the relevant length scales and time scales of adsorbed polymers by employing a coarse-grainedrepresentationof the polymer chain (in terms of discrete adsorption states) and stochastic dynamics. The dynamics of the adsorbed polymers are governed by a master equation. The transition probabilities that enter the master equation have been derived by applying transition-statetheory to the potential energy surfacethat describes the interactione of the PMMA segmenta with the aluminum surface. The potentialenergy surfacehas been obtained from previous quantum mechanical calculations for oligomers of PMMA near an aluminum surface. By incorporatingthe details of the segmentsurface interactions,we are able to examine the effects of tacticity on the dynamics of adsorbed chains. We find qualitative similarities and quantitativedifferences between the dynamics of isotactic and syndiotactic PMMA adsorbed on aluminum. The dynamics of PMMA chains share several features with the dynamics of glass-formingliquids. Both systems exhibit nonexponential relaration and non-Arrhenius temperature dependences of relaxation times. The physical sources of the glasslike dynamical behavior are a rugged energy surface and strong kinetic constraints. The microscopic basis for the kinetic constrainta is diecuesed in detail.

1. Introduction

The study of polymer adsorption at solid surfaces is motivated largely by the widespread technological application of polymer-solid interfaces. Colloidal dispersions, packaging of microelectronics components, composite materials, and biomedical devices are a few examples of applications that depend critically on the integrity of polymer-solid interfaces. Polymer-solid interfaces may be classified into two broad categoriesbased on the nature of the interactions between the polymer segments and the substrate. The first class of interfaces includes those in which the segment-surface interactions are weak and dispersive; the polymers are physisorbed at the interface. The second class of interfaces are those in which the segment-aurface interactions are strong and specific;the polymers are chemisorbedat the interface. Polypropylene on graphite is an example of the first class of interfaces; polymer-metal interfaces, such as poly(methy1methacrylate) (PMMA) on aluminum, are examples of the second class. In this work we focus specifically on the PMMAaluminum system. By analogy, we draw conclusions regarding the behavior of a wide range of strongly interacting polymer-solid interfaces. Physisorbed polymers at equilibrium are fairly well understood from both theoreticall-3 and experimental perspe~tives.~1~ In contrast, the study of strongly adsorbing polymers has been the subject of only recent attention. Chemisorbing polymers are markedly different from physisorbing polymers for several reasons. First, the segment-surface interactione are much stronger. Second, the chemisorptive interactions are highly specific. The chemisorbing functionalgroups adsorb preferentiallyover the more weakly interacting groups. It has been shown recentle7 that this preferential adsorption strongly perturbs the rotational conformational statistics from those observed in the bulk or in solution. Third, it has been demonstrated both theoretically7r8and experimentallyg that the adsorbed chains constitute a collection of nonAuthor to whom correspondence should be addreesed. 0024-929719312226-1120$04.00/0

equilibrium structures (a glassy interfacial layer) whose dynamical behavior is similar to that of glass-forming liquids (structural glaases).loJ1 Specifically, relaxation from nonequilibrium states is nonexponential, and the temperature dependence of relaxation t h e e does not follow an Arrhenius law. The trapping of adaorbed polymer c b in nonequilibrium states is a reault of the topography of the potential energy surface that characterizesthe interactionsbetween the polymer segmentsand the substrate. Studies of PMMA and aluminum6 have shown that the potential energy surface is " r ~ g g e d " ; ~ ~ J ~ that is, it contains many local minima and a broad distribution of activation barriers.I4 The presence of a glassy layer near the polymer-eolid interfacehas severaltheoreticaland practical implications. Of theoretical interest, the dynamicsof the polymer chains in the glassy layer are similar to those found in g h forming These similarities raiee intereating questionsregardingthe underlying physical proceasea that lead to glassy dynamics. Furthermore, the previous theoretical studies* have been performed with simple model polymers. The study of real systems, like PMMA on aluminum, with a rugged energy landscape presents a significant challenge. Overcoming this challenge should lead to insights regarding the microscopic origins of the observed dynamical behavior. From a practicalviewpoint, the existence of a glassy interfacial layer implies that the properties of the interface will depend on the history of preparation of the interface. While this has been recognized, the prediction of interfacial propertieaas a function of history for specificsystems remains an outstanding longterm goal. On the basis of the discussion above, it should be clear that strongly interacting polymer-solid interfaces have many special features whose theoretical analysis requires different methods than those used to study weakly physisorbed polymers. The interaction potentials cannot be constructed from simple, pa@iy additive potentials. Since the interactions are chemical in nature, interaction potentials must be derived from electronic structure calculations.6JS Once the interactions have been well 0 1993 American Chemical Society

Macromoleculee, Vol. 26, No.5, 1993 characterized, the equilibrium statistice of interfacial chains could be generated by the techniques that have been applied to weakly interacting systems.112 It has been argued, however, that the interfacial chains are not in equilibriumbut rather are constrained in nonequilibrium statea.7 Therefore,the study of theae systemsmust address not only static featurea but alsothe dynamicsof adsorption. Furthermore, the dynamics must be sampled from a wide range of initial conditions. In that way, the history dependenceof the interfacial structures can be addressed, and the appropriate nonequilibrium averages can be calculated. The final concem is the presence of very long time scalea. Sincethe interactionsof the polymer segments with the surface are so strong, the effective activation energy for conformationaltransitions or local desorption of segments is much larger than the thermal energy, ~ B T , at room temperature and below. The corresponding relaxation times can range from milliseconds to hours, and even beyond experimental time scales. Thus the important adsorption events occur over time scales that are impossible to probe using traditional molecular dynamics simulations; an alternative simulation procedure must be developed. The goal of this work is to develop a stochastic model of the dynamics of PMMA adsorption on an aluminum surface that can achieve the necessary goals for the study of strongly interacting interfaces set forth above. The stochastic model incorporates the details of the strength and specificity of the interaction potential; it follows the dynamics of the adsorption process explicitly; and it has the computational efficiencyto probe the experimentally relevant time scalesand average over many different initial conditions. By incorporatingthe details of the interaction potential, the model can also investigate the effect of the tacticity of PMMA on the dynamics of adsorption. The model is currently limited to isolated polymer chains, so it is most directly applicable to adsorption from dilute solutions. Nevertheless, we expect the physical phenomena revealed by the model to be important for interacting chains and for adsorption from concentrated solutions. This point will be addressed in more detail later. The format of the paper is as follows. In section 2 we review the relevant features of the PMMA-aluminum system and formulate the stochastic model for the dynamics of strongly adsorbed polymer chains. In section 3 we describe the potential energy function used to model the interactions of PMMA with aluminum. The computational methods used in the study are given in section 4. In section 5 the results of the stochastic simulations of adsorbed chains are presented and discussed. The resulta are interpreted in view of theories developed for other glassy materials and compared with recent experimental work on polymer adsorption. Concluding remarks are offered in section 6. 2. Model Development and Theory 2.1. Overview. Our goal is to construct a model for the dynamics of PMMA chains adsorbed on aluminum surfaces. First we review the important features of the PMMA-aluminum hterface. The interactions between segments of PMMA and an aluminum surface are strong and specific. In particular,the interaction energy between PMMA segments and an aluminum surface is highly dependent on the orientation and configuration in which segmenta approach the surface.6 The orientation dependence arks becausethe methacrylate sidegroup of PMMA a carbonyl group and an ester oxygen atom that conmay chemisorb at the surface, while the alkyl backbone

Dynamics of PMMA Chains Adsorbed on Al Surfaces 1121 of the polymer cannot interact so strongly. Binding energies for chemisorption of the carbonyl group range from approximately 60 to 100 kJ mol-', while the interaction of the ester oxygen atom is weaker, with binding energies ranging from approximately 20 to 50 kJ mol-'. In orientationsin which either chemisorbingfunctionalgroup can approach the surface, there is an activation energy to adsorption of approximately 50 kJ mol-'. Activation energies for the complete desorption of segments range from 100 to 150 kJ mol-', while activation energies for conformationaltransitions between different chemisorbed conformations are much lower, approximately50 kJ mol-'. By considering the strength of the interactions, we see immediatelythat conventionalmolecular dynamics (MD) simulations cannot probe the time scales relevant to the PMMA-aluminum system. For example, from the activation energies listed above,we can estimate that the time scale for conformationalrelaxation of adsorbed segments at 300 K will be on the order of lP3s. Furthermore, the existence of a glassy interfacial layer' requires the calculation of nonequilibrium averages over the initial configuration space. As a result, simulationsmust explore many different realizations of the adsorption process subject to a wide range of initial conditions. Given that a single MD simulation of polymer and solvent can realisticallyprobe only 10-88, it is clear that the application of MD simulations to this problem is an impossible task. Recently, Chakraborty and Adriani have proposed a kinetic Ising modelsJ6J7 to address the shortcomings of MI) Simulations. The model does not address PMMAaluminum interfaces specifically but rather considers a generic representation of a strongly adsorbing polymer which contains two functional group per segment ('stickers") that can bind strongly to a surface. This polymer architecture is a schematic representation of PMMA, in which both the carbonyl group and ester oxygen atom may chemisorb on aluminum.6 We note that other simple chain architectures have been considered in ref 8. The fundamental premise of the kinetic Ising model is that once the polymer is adsorbed in an initial nonequilibrium conformation, the only relevant information is whether each sticker is adsorbed or desorbed. In other words, the explicit atomic degrees of freedom of the polymer and solvent are discarded, and each sticker is mapped onto a two-state system, or an Ising spin variable. The spin variablesthen evolve in time according to dynamicalrules that were chosen based on physical considerations that reflect the polymer chain connectivity. The dynamical rules account for local torsional strain, steric hindrance, and entropic constraintaimposed by the surface. By using such a description, the model revealed that the dynamics of strongly adsorbed polymer chains were akin to those of glass-formingliquids. For example,the simulationsreveal that the interfacial chains are locked into nonequilibrium conformations at low temperatures. At intermediate temperatures, the chains can relax to equilibrium in reasonable times, but the dynamics are very sluggish, showing similarities to bulk glass-forming liquids. In particular, the relaxation to equilibrium was found to be nonexponential; the dynamics could be described with a stretched-exponential, or Kohlrausch-Williams-Watts (KWW), function: q(t) = ~ x P [ - ( ~ I T ~ ~ ~ ) ~ I

(1)

where q(t) is a nondimensional measure of relaxation to equilibrium, and B and TKWW are adjustable parameters. Also,the temperature dependence of the relaxationtimes, 7 ,did not follow an Arrhenius law but showed a divergence

1122 Shaffer and Chakraborty

that was described by the Vogel-Fulcher equation:

The parameter TW physically corresponds to a temperature at which the relaxation time diverges or exceeds observational time scales due to dynamical falling out of equilibrium. The Vogel-Fulcher equation, which predicts a divergingrelaxation time, should be contrasted with the Arrhenius form d T ) = 70 exp(E,/kBT)

(3) which predicts a finite relaxation time for any finite temperature. In eq 3,E k is an effective activation energy and so is an intrinsic microscopic relaxation time. To examine the adsorption of PMMA on aluminum, we take an approach similar to the kinetic Ising models of adsorbed chain dynamics proposed in ref 8: we adopt a "coarse-grained" stochastic model, one that follows the dynamics of a reduced set of degrees of freedom. The details of the model and the simulationtechnique, however, are quite different from ref 8 because we incorporate the features of the segment-surface interactions that are specific to the PMMA-aluminum interface. The formulation to be developed in this paper has three important components: (1)Each monomer of the polymer chain is classified into one of four discrete states based on the sorption state of the two chemisorbing functional groups. Information regarding the sorption states is obtained from an analysis of the PMMA-aluminum potential energy surfaceobtained from quantum mechanical calculations.6 (2) The time evolution of the polymer configuration is modeled as a stochastic process governed by a master equation. The transition probabilities that enter the master equation play a crucial role in determining the dynamical behavior. In simple treatments: phenomenological rules were invented to construct a model for the transition probabilities. In our model, the transition probabilities are obtained from the application of transition-state theory to the detailed potential energy surface that characterizes the PMMA-aluminum system. (3) Particular realizations of the adsorption dynamics are obtained by numerical solution of the governing master equation. We will now describe each part of the formulation in turn. 2.2. Discrete-StateClamifleation. It is known from MD simulations7t8and experiments1*that the first step of adsorption from solution is fast. It is the subsequent relaxation of the adsorbed chains Erom this initial nonequilibriumstatethat is of interest here. Oncethe PMMA chain is in a nonequilibrium conformation with a few segments adsorbed, the important events during the relaxation to equilibrium will be the adsorption and desorption of the chemisorbing functional groups, the carbonyl group and the ester oxygen.6 It is appropriate, then, to take acoarse-grained representationof the polymer chain in which an integer label, Q, is associated with each monomer of the chain according to the sorption state of the carbonyl group and the ester oxygen atom. Specifically, each monomer can be present in one of four discrete states: (1) the carbonyl group adsorbed, ester oxygen desorbed; (2) carbonyl group desorbed, ester oxygen adsorbed; (3) both functional groups adsorbed; and (4) both functionalgroupdesorbed. (Polymersegmentswith both functionalgroupsadsorbed will frequentlybe referred to as "strongly adsorbed"; segmenta with both functional groups desorbed w i l l often be called 'free" segments.) A particular functional group is considered to be adsorbed

hhm"k?CUk?8,

Vol. 26, No.5, 1993

if it resides cloeer than 4 A to the surface; thie choice is based on the location of the attractive part of the potential energy surface obtained in ref 6. It is important to note that each discrete state is comprised of many stable conformations (local minima on the potential energy surface) that differ only in degrees of freedom that are irrelevant for the chemisorptionof segments. The torsion angle of a weakly interacting pendant methyl group is an example of an irrelevant degree of freedom. 2.3. Transition Probabilities. Given the coarsegrained descriptionof the polymer configuration,the heart of the stochastic model is the specification of the rate constants, or transition probabaties, for transitions between the various discrete states. As noted above, in the kinetic Ising model of ref 8 the transition rates were poatulated based on physical considerationsof the polymer chain connectivity. Here, however, we will calculate rate constants directly from the potential energy surface by applying transition-state theory. With the discrete-state claasiication given above,the configurationof an adsorbed polymer chain of N segmenta is given by the set of N integers (81,Q2, ..., QN)= Q. The configuration space of the adsorbed polymer is reduced to the 4* possible arrangements of the monomer state variables. The fundamental quantities that we seek are the rate constants R(Q-Q), the transition probabilities per unit time for the polymer chain to undergoa changefrom configuration Q to confiiation Q'. Oncethese transition probabilitiea are specified, the statistical behavior of the adsorption process is completelydetermined by the master equation19

whereP(Q,t) is the probability of the polymer chain being in confiiation Q at time t. We make the assumption that only one monomer at a time changes ita sorption state. The transition probabilities above can then be written R(Q+'iIQ), the probability per unit time for monomer i to make a transition from state Qi to state Q'i, given that the polymer chain is in configuration Q. The rate constants for any given monomer to change ita sorption state depend on the current state of the monomer becamethe segment-ilurfaca potential is orientation and configuration dependent. The rate constanta also depend on the states of neighboring monomersbecauseof torsionalstrain and steric hindrance; this is a local effect mediated through the torsional and nonbonded interaction potentials. There is also a n o n l d effect on the rate constants from the configurational entropy change that accompanies the adsorption or desorption of polymer segments. To account for these separate effects, we factor the rate constant into local and nonlocal terms as follows: R(Q,4'AQ) RL(Qi+Q'ilQj-1,Qi+l) &,(Qi*Q'ilQ) (5) The factorization above is appealing because it separates the configurational entropy, which is generic to all polymer-aolid interfaces, from the local effects that are unique to each polymer-aolid system coneidered. We further assume that the local effects on the adsorption and desorption ratee of a given monomer can be represented by accounting for the in~tantaneousstatee of ita two nearest neighbors only. As a result, the qunntitiea RL(Ql-qi)Qi-l,Qi+1) can be derived from the potential energy surface of a trimer of PMMA interacting with an aluminum surface. Note that the effect of the tacticity of the PMMA chains is incorporated explicitly in this

Macromolecules, Vol. 26, No. 5, 1993 treatment. For the isotactic chain, the transition probabilities are calculated using a PMMA trimer consisting of two meso diads; for the syndiotactic chain, they are calculated usinga trimer consisting of two racemic diads. The calculation of the nonlocal rate factoreRm(Q,+Q'ilQ) requires the treatment of the configurational entropy of polymer chains near surfaces. Below we describe our treatment of both contributions to the transition probabilities. Discrete-StateTransition-StateTheory. As noted earlier, we make the assumption that the local effecte on the transition rates of a PMMA monomer can be captured by considering only the instantaneous states of its nearest neighbors. Since only three monomers are involved, we can determineall localrateconstants,RL(Q'~~Qi-i,Qi+l), by applyingtransition-statetheory (TST)to the potential energy surface of a PMMA trimer interacting with aluminum. In traditional applications of TST, the configuration space is divided into regions, called 'mithat contain a single local minimum on the crostate~~, potential energy surface.20*21 The microstates are separated by dividing surfaces that contain transition states, first-order saddle points on the potential energy surface. Using one of several theoretical prescriptions,22a firstorder rate constant can be calculated for transitions between microstates. Here, however, within the coarsegrained classification of polymer configurations into discrete sorption states, each discrete state will encompass many microstates. Furthermore, there will be many conformationaltransitions within a discrete state, as well as many possible transition paths between discrete states. Given a complete catalog of potential energy minima, transition states, and the rate constants for jumps between minima, a method is required to construct the overall transition probabilities, RL, between the discrete states by appropriately weighting the individual rate Constants. The construction of the overall rate constants involves an argumentbased on the separationof time scalesbetween conformationaltransitions of weactive side groups and the activated adsorption and desorption of chemisorbing functionalgroups. Considera large number of microstates that have been classified into one of two 'macrostates", based on a clear physical distinction. For example, for a PMMA trimer near an aluminum surface, the first macrostate may contain all locally stable conformations in which only the carbonyl group of the middle segment is chemisorbed to the surface;the second macrostate may contain all stable conformations in which only the ester oxygen atom of the middle segment is chemisorbed to the surface. Supposenow that all reaction paths for transitions between the individual potential energy minima (microstates) are known. The reaction paths can be distinguished as to whether or not they cross a macrostate boundary, that is, whether the microstates involved (the initial and final points in configuration space) belong to the same macrostate or to two different macrostates. Reaction paths that connect two microstates within the same macrostate will generally correspond to a simple conformational transition of a weakly interacting side group. For such transitions, the activation energies are relatively low. Conversely, paths that connect two microstates in different macrostates will involve the desorption of a chemisorbed functional group or the adsorption of a different group. The corresponding activation barriers are very high compared to the activation energiesof simple conformationaltransitions. As a result, the rata constante for transitions between two microstates within the same macrostate will be orders of magnitude

Dynamics of PMMA Chains Adsorbed on Al Surfaces 1123

higher than rate constants for transitions between macrostates. The discrepancy in the rate constants leads to a separation of time scales between the two claeses of events. Given the separation of time scales as outlined above, it is a goad approximation to treat the microstates within a given macrostate as locally equilibrated during the time intervals between jumps from one macrostate to another. In other words, once the system makes a transition from one macrostate to another, the microstateswithin the final macrostatequickly become populated with the probability distribution Pf = zA-'leXp[-V(x)/kBfl dx

(6)

In the equation above,pf is the probability of microstate i in macrostate A being occupied when the system is restricted to macroatate A, x repreaents the configurational degrees of freedom (atomic Cartesian coordinates), V(x) is the potential energy function, and the integration is limited to the region of confiiation space belonging to microstate i. The normalization factor ZA is the locul configurational partition function for macrostate A:

ZA = JAexp[-V(x)/kBT] dx (7) where the integration for ZAis over the configurationspace of all microstates that have been grouped into macrostate A.23 Clearly eq 6 applies only on time scales shorter than the average time required to make transitions to other macrostates. Sincethe microstate populationsare rapidly redistributed before the system has a chance to jump to another macrostate, the individual rate constants for transitions out of the macrostate must be weighted according to eq 6. The effective rate constant for transitions between macrostate A and another macrostate B is then calculated by

whereRA-Bis the overalleffectiverate constant (transition probability per unit time) for transitions from macrostate A to macrostate B. The sum is over all individualreaction paths that connect microstates in macrostate A with microstates in macrostate B; the ki-j represent the rate constants for the individual transitions between microstates. By summing over all paths between the macrostates, the effective rate constants account for entropic bottlenecks in configurationspace in addition to potential energy barriers for individual reaction paths. In effect,the potential energysurfaceof the full configuration space has been mapped onto a free energy surface of much lower dimensionality.23 In the Appendix it is shown that the effective rate constants defined by eq 8 satisfy the condition of detailed balance for the macrostate populations. Configurational Entropy. The entropic constraints associated with changing the loop, train, and tail distributions by adsorption or desorption are represented through the rate constant RNL(QT+Q'ilQ) in eq 5. In general, RNLcan depend on the entire configuration of the polymer chain, as well as the current and target states of the specified segment. The change in entropy for a given event can be estimated by counting the number of configurations, W, available in loops and tails, and using the formula S = kB In W. The following expressions can be derived for random walks on lattices confined to one side of a surface plane:4*241*5

Macromolecules, Vol. 26, No. 6, 1993

1124 Shaffer and Chakraborty

by

where WL and WT are the number of conformations available to a loop and tail, respectively, comprised of n - 1segments. By using expFessions for random walks in one dimension, we implicitly account for the fact that, in the random-walk model, the interface does not affect the configurational freedom in the two coordinate directions parallel to it. The factors n-3/2 and n-1/2in eqs 9 and 10 indicatethe loss of configurationalentropy aesociatedwith forming loops and tails. They also show that one long loop is entropically favored over two shorter loops. For transitions between different adsorbed states of the same segment, RNL is obviously unity. For transitions that involve adsorption or desorption, we must work out transition rates for adsorption-desorption events that satisfy detailed balance. The ratio of rate constants must be equal to exp(AS/kB), where AS is the entropy change that accompanies the event. We choose RNL= exp(AS/ kg) for adsorptioneventsand RNL= 1for the corresponding desorption events. Thus the rate constant for adsorbing aloop or fail segmentis lowered due to the negative entropy change. 2.4. Time Evolution. To study the dynamics of polymer adsorption, we must obtain solutions to the governing master equation, eq 4, for the time evolution of the probability P(Q,t)of each possible configuration. The high dimensionality of the problem and the long-range correlations introduced by the entropic constraints preclude an analytical solution. Instead, we perform stochastic simulationsof the adsorption process starting from many different initial conditionsand accumulateaverages. The stochastic simulations employ the theory of pure Markovjump p r o c e ~ s e s . To ~ *use ~ this theory, we assume that, for each configuration of the polymer chain, all possible adsorption and desorption events (the 'elementaryevents") are independent Poisson processes. That is, each poesible event Q L d Q ' i has a first-order rate constant R = R(Ql+'ilQ) associated with it, as discussed above. The time interval between events is an exponentially distributed random variable:

F ( t ) = R exp(-Rt) (11) where F(t) is the probability density for the distribution of time intervals between events. The average time between events, ( t ) , is calculated as (t)=

JOmtF ( t ) dt

(12)

and is equal to the inverse of the rate constant, 1/R. Several properties of exponentiallydistributed random variables% have been used to construct the simulation algorithm. Consider n distinct stochastic eventsgoverned by Poisson statistics accordingto eq 11,with rate constants R1, Rz, Rn, that occur in time intervals tl, tz, ...,t,. The time interval for the first event to occur, 71 = min(t1, tz, ...,t,),isalaoanexponentiallydistributedrandomvariable with an overall rate constant, p , equal to the sum of all the individual rate constants:

...,

PI = R,/P (14) so that the total probability is normalized. It can also be shown that the stochasticproceseconsistingof the n events is Markovian.26 Using the propertiesof the Markovjump processdefined above, the implementation of the simulation algorithm is The initial configuration of the adsorbed polymer is generated randomly. The rate constants (transition probabilities) for each poesible adsorption and desorptionevent are calculated based on the current chain configuration,as outlined above. The rate constants are then summed according to eq 13, and the time interval 71 to the first event is sampled from the probability distribution of eq 11with rate constant p. The time 71 is then added to the simulation clock. Finally, the actual event to occur at this time step is chosen from the probability distribution of eq 14. Once the time and event have been determined stochastically, the polymer confiiation is updated and a new step in the simulation begins. This simulation procedure, in which the time step as well as the sorption events are stochastic variables, has the great advantage that one event occurs at each step in the simulation. We never have to wait for an event to occur, but rather the time step os the simulation adjusts to the underlying time scale of the physical process. This feature is especially valuable in simulating the dynamics of strongly adsorbed polymers. The first phase of the adsorption process is very faet,8J* but the ultimate relaxation to equilibrium can be extremelyslow,and even surpaes experimentaltime scales. Thus, the adaptivetime step is essential in following the relaxation to equilibrium at low temperatures.

3. Potential Energy Function To have a realistic model of adeorbed chain dynamics at PMMA-aluminum interfaces, the transition probabilities in the stochastic model must incorporate the physics and chemistry of the PMMA-aluminum system. As discuseed in section 2, the quantities RL(Qld@dQj-l,Qj+l) can be derived from the potential energysurfaceof a trimer of PMMA interacting with an aluminum surface. In this section, we present the potential energy function that models the intramolecular potential for a PMMA trimer and the potential that describes the interactions of a PMMA trimer with an aluminum surface. The potential energy function, V, for the PMMAaluminum system is the sum of the intramolecular potential, P b * , for the PMMA trimer alone, and the surface potential, V", that describes the interaction of the PMMA trimer with the aluminum surface. The intramolecular potential has been adapted from previous work of Vacatello and F10ry.~For computational efficiency, all pendant methyl groups (-CHs) and backbone methylene groups (-CH2-) are treated as united atoms. Furthermore, all bond lengtha and bond angles are fixed. Following ref 29, the intradiad bond angle (denoted by f in ref 29) isfiied at 124O;the interdiad bond angle (denoted by 7 in ref 29) is fued at 1 1 1 O . All other bond angles and all bond lengtha are fmed at their minimum energy values. With the simplificationof constant bond lengthsand bond angles, the intramolecular potential ia written 88 follow:

n

P =

p,

(13)

,=I

The probability that the ith event will occur first is given

The first term is a sum over all torsion angles &; Vitorie

Macromolecules, Vol. 26, No.5, 1993

Dynamics of PMMA Chains Adsorbed on Al Surfaces 1125

Table I Atomic Data for Calculating Le"Jones Parameters from the Mater-Kirkwood Formula van der Waals effective no. polarizability, atom radius. A of electrons A3 1.8 5 0.93 C C (carbonyl) 1.8 5 1.23 0 (carbonyl) 1.6 7 0.84 0 (ester) 1.6 7 0.70 CH2 or CHf 2.0 7 1.77 0 Sundararajan, P. R.; Flory, P. J. J. Am. Chem. SOC.1974, 96, 5025.

Table I1 Lennard-Jon- Parameters between 'Nonbinding" Atoms in PMMA and Aluminum for Use in eas 16 and 17 ~~

atom

e,,

C C (carbonyl) CHI or CHI

kJ mol-' 7.29 9.13 5.37

ul, A

2.88 2.88 3.24

the potential specific for each distinct torsion angle given in ref 29. The second term is a sum over all nonbonded pairs. The distance between atoms i and j is denoted by rij; partial charges are denoted by qi and qj. LennardJones potentials, with parameters Aij and Bij, act between all nonbonded pairs; the Coulomb potential, represented by the last term, acts only between the atoms in the methacrylate side groups that carry significant partial charges. The parameters for the torsional potentials, partial charges, and the dielectric constant, c, are taken directly from ref 29. The Lennard-Jones parameters Bij are calculated from the Slater-Kirkwood formula, with the atomic parameters given in Table I; the set of Aij are calculated such that the Lennard-Jones potential between groups i and j is a minimum at the sum of their van der Waals radii. The surface potential, V8&, has been derived from a previousquantum mechanical stud9 of PMMA oligomers near an aluminum surface in which the carbonyl bond was allowed to relax upon chemisorption. An empirical force field was presented in ref 6; here, however, we have taken the same interaction energiesas those of ref 6 and fit them to a slightly different functional form. For the purpose of the interaction potential, the atoms in PMMA are divided into two groups: the two oxygen atoms of each PMMA monomer are considered to be "binding"; all other atoms are "nonbinding". The nonbinding atoms interact with the surface through a purely repulsive, truncated and shifted Lennard-Jones potential?

] for zi I (i) ui (16)

2 5 112

-(-) 32

116

V y = 0 for z i > ( i ) l l 6 U i In the equationsabove, Vp..ie the potentialfor nonbinding atom i; Zj is the distance of atom i from the aluminum surfaw p ~ isl the number density of aluminum atoms, 0.0603A-3; ~i and ci are Le& Jonee parameters between the nonbindingatoms of PMMA and the aluminum atoms in the substrate. The Lennard-Jones parameters for use in eqs 16 and 17 are given in Table 11. In calculating the parameters from the Slater-Kirkwood formula, the polarizability of aluminum was taken from the work of de Heer et al.31 on large aluminum clusters.

Table 111 Coefficients for the "Binding"Potentials (Equations 18-20) Shared Parameters A = 50.0 kJ mol-' zb p 2.0A A = 1.1A-1 Ehter Oxygen Parameters, kJ mol-' B = 157.0 C = 87.0 Carbonyl Oxygen Parameters, kJ mol-' bo 148.7 70 114.3 8' = 47.4 y1 = -28.4 82 = -4.9 yz = -49.9

The chemisorptionof PMMA to aluminum is described by using a modified Morse potential for the binding atoms: eXp[-BA(Zi - zb)] - Bj eXp[-X(Zi - zb)] + ciA2(zi- Zb)' exp[-A(zi - 291 (18) where Vibindis the potential for binding atom i. The fmt two terms of eq 18form the standard Morse potential; the third term represents the activation energy for chemisorption. The values of A, A, and Zb in Table 111;they are the same for both the carbonyl and ester oxygen atoms. Table 111 also contains the values of the coefficientsB and C for the ester oxygen atom. The orientation dependence of the binding and activation energies is incorporated in the coefficients B and C for the carbonyl oxygen atom. The coefficients are not constant but depend on the orientation of the carbonyl group as follows:

VPd

Aj

B = Po +

cos e + p2 cos2e

(19)

c = yo+ yl COS e + yz cos2e

(20)

where 8 is the angle between the surface normal vector and the vector drawn from the carbonyl oxygen atom to the carbonyl atom. (The angle 8 = 0 if the carbonyl group is perpendicular to the surface, with the carbonyl oxygen atom closer to the surfacethanthe carbonylcarbon atom.) Table I11 contains the values of the constants used in eqs 19 and 20. The parameters in eqs 18-20 were fit to the results of the quantum mechanical calculations reported in ref 6. (Prior to the fitting procedure, the parameters in eqs 16 and 17 were fixed according to the values in Table I1 and were not adjusted.) Finally, the total interaction potential between the PMMA trimer and the aluminum surface., Pd, is simplythe sum of the individual contributions from each Vponand Vibind* 4. Computational Methods A prerequisite for the application of TST is the location of all relevant extrema (minima and first-order saddle points) and the associated reaction paths on the potential energy ~urface.3~ We accomplish this task in two steps. First, we perform a conformationalsearch that locates the saddle points on the potential energy surface. Then, we trace the reaction paths from transition states to potential energy minima by following steepest-deecent paths. The conformationalsearch for saddlepoints ie bawlon the "usage-direded"searching algorithm deecribed in detail by Chang et al.= for the location of potential energy minima in organic molecules. The only modification in our procedure ie to search for firstorder saddle points instead of minima. Starting from a given point on the potential energy surface (generated according to the algorithm given in ref 33), we use the rational function optimization (RFO)method%@to stepalong the potential energy surface to the "nearest" first-order saddle point. This transition state is then compared to the existing

Macromolecules, Vol. 26, No.5, 1993

1126 Shaffer and Chakraborty

database of previously located transition states. The iterative search procedure is continued until no new transition states are located in 10 OOO attempts. The RFO method requires that the Hessian matrix of second derivatives of energy with respect to coordinates be evaluated and diagonalized. This is computationally demanding but still feasible, with the number of degrees of freedom in the PMMA trimer. After the transition states have been located and cataloged,the reaction paths and potential energy minima are obtained by following steepest-descent paths in massweighted coordinate from the transition states down to the associated minima. The steepest-descent paths are initiated along the single direction of negative curvature at the transition state as indicated by the Hessian eigenvectorassociatedwith the singlenegative eigenvalue. The first steepest-descent path originates parallel to this eigenvector; the second originate antiparallel to this eigenvector. The potential energy minima that are connected by a reaction path that goes through the transition state are then identified as the terminal points of the two steepest-descent paths. The calculation of the transition probabilities RL(Qc+Q'ilQi-ltQi+l) calle for the evaluation of the configurationalintegrals defined in eqs 6 and 7. In principle, Monte Carlo techniques could be used, but the high dimensionality of the potential energy surface and the number of minima (greater than 30 OOO) make the computation unfeasible. For simplicity,the local equilibrium probabilities pf were calculated as pf =' 2 ;

eXp(-V,/kBT)

(21)

where V, is the potential energy at the minimum that defines microstate i. The local partition function, ZA,is given by

where the sum is taken only over microstatesin macrostate A. Given a pair of potential energy minima and the transition state that connects them, the first-order rate constants were calculated simply as

where k i , is the rate constant for transitions from minimum (microstate) i to minimum j , and vice versa; Vm is the potential energy of the transition state; V, and Vj are the potential energiesof the minima. The prefactor ko was chosento be 10'2 s-l to reflect the order of magnitude of theattempt or collisionfrequencyfor activatedprocesses in solution.% The rate constants could have been calculated more accurately by applying TST for multidimensional, multistate systems and accounting for dynamical corredi01~,3~~38 but the number of distinct reaction paths (on the order of 105) prohibited the more detailed calculations. By using only the potential energy at the minimum in eqs 21 and 22, we are implicitly assuming that the entropic contribution to the equilibrium probabilities is comparable for all stable conformations. Similarly, by using a constant prefactor in eqs 23 and 24, we aesume that the configurational entropies of the transition states are comparable to those of the stable states.

0101

-400

-200

0

200

Energy (W mol-')

Figure 1. Distributionof potential energy minima for a PMMA trimer interacting with an aluminum surface. The solid line corresponds to the mesemeso trimer (model compound for isotactic PMMA chains); the dotted line corresponds to the racemic-racemic trimer (model compound for syndiotactic PMMAchains).

5. Results and Discussion

6.1. Energy Distributions. Before proceeding to the discussion of the dynamics of PMMA chains adsorbed on aluminum, we discuss two features of the PMMAaluminum potential energy surface: the distribution of potential energy minima and the distribution of activation barriers to conformational transitions and adsorptiondesorption events. These energy distributions illustrate the structure of the potential energy surface at the 'local" level of PMMA trimers. The potential energy surface at this level determines the transition probabilities RL(Qh+Q'ilQr-1,Qi+l). The conformational searches for reaction paths for the isotactic and syndiotactic trimers of PMMA near an aluminumsurfacediscoveredmore than 30000 minima and 1OOOOO transition s t a t e for each stereoisomer. There are still more extrema that were not cataloged because of limitations in computational resources. The main sourceof difficultyis that the efficiency of the conformational search, defined as the probability of finding a new saddle point on a given trial, decreases rapidly as the number of trials increwes. Neverthela, the large database of extrema should provide a good statistical description of the full set of minima and transition states. Furthermore, because individual transitions are weighted by their probability of occurrence accordingto eq 8, only a Statisticalsamplingof the potential energy surface is required to construct the coarse-grained rate constants. The distributions of potential energy minima for the isotactic and syndiotactic trimers are presented in Figure 1. The figure shows the fraction of minima as a function of the potential energy of the combined PMMA trimeraluminum system. The zero of energy in each case correspondsto the most stable conformationof the PMMA trimer in vacuum. Note that the distribution of energy minima is very broad, both for adsorbed (Vi < 0) and desorbed (V, 2 0) trimers. The minima with the lowest energies (Vi -400 kJ mol-') are those in which all three monomers have chemisorbed to the surface with both the carbonyl group end ester oxygen atom. Only a small fraction (0.006) of the conformations,however,dow strong chemisorption to the aluminum surface; most conformatione in which all three monomers chemisorb have higher energies. In these confiiations,a high degreeof torsional f~

Macromolecules, Vol. 26, No.5, 1993

0

50

100

150

Dynamics of PMMA Chains Adsorbed on AI Surfacea 1127

200

250

Activation energy (kJ mol-’)

Figure 2. Distribution of activation energies for conformational and adsorption-desorption transitions of a PMMA trimer interactingwithan aluminum surface. The solid line corresponds to the meso-mew trimer (modelcompound for isotactic PMMA chains);the dotted line correspondeto the racemieracemic trimer (model compound for syndiotactic PMMA chains).

strain or steric crowding occurs to accommodate the chemisorption. As a result of the wide energydistribution, the effective free energy of adsorption per monomer and the adsorbed chain conformations will be temperature dependent. The strongly bound confiiations will dominate at low temperatures, while the more weakly bound configurations will become populated at higher temperatures. Figure 2 shows the distributions of activation barriers for transitions between potential energy minima of the PMMA-aluminum system. These events include conformational transitions between different adsorbed configurationsand transitionsbetween adsorbed and desorbed configurations. The distributions for the two trimer tacticities are very similar. The large fraction of events with activation energies close to zero are transitions between very s i m i i adsorbed configurations. The sharp peak in the distributions for both stereoisomersbetween 15 and 20 kJ mol-’ corresponds to conformational transitions of backbone and pendant group torsion angles in desorbed monomers. The fairlywide clustersof activation energies centered around 50 and 80 kJ mol-’ are barriers for conformational rearrangement of chemisorbed segmenta and for the activated adsorption of segments. Finally,the longtailof barriers at higher activation energies correspond to transitions with a greater degree of steric hindrance than the lower energy transitions. Note that the dietributions in Figure 2 are similar, but not identical, for the two stereoisomers. The effects of these wide energy dwtributions (both for minima and for activation barriers) on the dynamical properties of adsorbed isotactic and syndiotactic PMMA chains are discussed below. 6.2. Dynamics. Below we present the results of the stochastic model for the dynamics of PMMA chains adsorbed on aluminum surfaces. A qualitative discussion is given firet; a quantitative characterization follows. We have examined both isotactic and syndiotactic polymers over a wide range of temperatures. Results for the two tacticitiea share many qualitative features, although there are quantitative differences. The resulta for the two tacticities are presented together to facilitate the comparison between the two chain architectures. In all cases for which data are presented, the number of segments,N, is 1OOO. The effect of chain length has not been inves-

tigated systematically. Preliminary investigations revealed only a minor effect of chain length on the results discussed below, indicating that the infinite chain length limit is approached at N =: 1000. In the discussions that follow, we present results for a wide range of temperatures, many of which cannot be realized experimentally. However, the presentation below will show that the dynamical behavior of the PMMAaluminum system must be studied over a wide range of temperatures to elucidate the physical mechanisms that determine the behavior at room temperature. Furthermore, the most important quantities for the dynamics of adsorbed chains are the activation energies for conformational transitions and adsorption-desorption events divided by kBT. Spanning a wide range of temperatures is equivalent to spanning a wide range of activation energies at a fixed temperature. Thus,the dynamical behavior of the PMMA-aluminum system at high temperatures may show qualitative similarities to more weakly interacting polymer-solid system at room temperature. For emphasis throughout the text below, when we quote a given temperature, we also specify the quantity EJkBT, where Ea = 50 kJ mol-’ is a representative activation energy both for the adsorption of PMMA segments and for conformational transitions between different adsorbed configurations. This activation energy is the most appropriate one to renormalize the temperature scale because the two most important eventsin the dynamicsof adsorbed PMMA are the adsorption of segments and the subsequent conformational relaxation of adsorbed segments. Qualitative Description. Here we present a qualitative view of the adsorption process for both the isotactic and syndiotactic chains at a low temperature, 300 K (Ed kBT = 20), and a high temperature, 2000 K (EdkBT= 3). The dynamicsof adsorbed polymer chains are followed by accumulating statistics for the time evolution of several quantities of interest: the number of polymer segmenta in each discrete sorption state, the number of individual adsorption-desorption events occurring in a given time interval, and the average train length of the adsorbed chains. These quantities are plotted in Figures 3-5. Because of the wide range of time scales, all time axes are logarithmic. The data presented in Figures 3-5 are averages over 100 realizations of the adsorption procees. The initial adsorbed configurations were generated by randomly placing 5 % of the segments in each of the three adsorbed states; the remaining 85% were then free segments in loops and tails. Consider first the adsorption of an isotactic chain at 300 K. A striking feature of the adsorption processs is the presence of a wide range of time scales. Several distinct time regimes are apparent from the evolution of the sorption state populations, shown in Figure 3a, and the distribution of individual adsorption-deaorption events, 8, the number shown in Figure 4a. At times less than of free segmentsremains nearly constant while the number of segmenta with only one functional group adsorbed decreases. Thus, most eventa in this period involve conformational transitions among segments that were initially adsorbed; typically, the segments with only one functional group adsorbed are able to adsorb the second group. After lo” s, the number of free segments falls drastically; concomitantly, the number of strongly adsorbed segments (those with both functional groups adsorbed) and the average train length, shown in Figure Sa, grow rapidly. In this time regime, the loops and tails are adsorbing to the surface.

Macromolecules, Vol. 26,No.6, 1993

1128 Shaffer and Chakraborty

Time (s) 1000

Time (s)

- 1 l 800

o

o

10'"

0

q

10'0

Time (s)

Time (s)

Figure 3. Time evolution of the discrete-state populations for PMMA chains of 1000 segmenta adsorbed on an aluminum surface: (a) isotactic chain at 300 K; (b) syndiotactic chain at 300 K; (c) isotactic chain at 2000 K (d) syndiotactic chain at 2OOO K. At 900 K, EJkBT = 20; at 2000 K, EJkBT = 3. See the text for the discuseion of renormalizing the temperature scale with the activation energy E.. Dotted line: carbonylgroup adsorbed,ester oxygen atom desorbed. Dot-dashed line: ester oxygen atom adsorbed,carbonyl group desorbed. Solid line: both adsorbed. Dashed line: neither adsorbed.

Closer examination of the adsorption simulations and and the transition probabilities RL(Q'-+Q'jlQi-1,8'+1) R n ( Q l 4 Q ' i l Q ) hae revealed that the adsorption of loops and tails occurs through a cooperativedynamical process. Qualitatively similar behavior has been seen in previous kinetic Ising models of polymer adsorption.8 The cooperativity arises because the adsorptionrate of an individual segment is greatly enhanced if one of ita nearest neighbors is already adsorbed. In other words, the adsorption of a given segment is kinetically constrained until one of ita neighbore adsorbs. For example, in the isotactic chain at 300 K,the rate conatantRLfor the adsorptionof a desorbed segment (into a configuration with both chemisorbing functional groups adsorbed) is 1.72 X 106 8-l when one neighbor is strongly adsorbed and the other neighbor is desorbed, the rate constantis 0.630s-lwhen both neighbors are desorbed. Both local and nonlocal effects account for this difference of 6 orders of magnitude. The important factors at the local level include not only torsional strain and nonbonded interactionsbetween monomers but also the activation barrier to chemisorption of the strongly interacting functional groups. Once one aegment is adsorbed to the surface, it can help to upull"ita neighbor over the activation barrier. (One could imagine that, for a polymer with a bulky side group, like polystyrene, en adsorbed segment could actually hinder the adsorption of

ita neighbors because of blocking by the sidegroup.) There is also a n o n l d effect due to thechange in configurational entropy that accompanies the adsorption of a segment in a loop. It can be seen from eqs 9 and 10 that the entropic penalty for adsorbinga loop or tail segment is least for the anchoringsegmentaof the loop or tail(thesegmentaneareat to the surface and adjacent to a train). Thus, the nonlocal rate constant Rn(QB+'&) will be grsatest for the anchoringsegments of loops and tails. The facilitation of the adsorption process through nearest-neighbor interactions was postulated in the kinetic Ieing modela of strongly interacting polymera;8 here that postulate is verified directly from the potential energy surface. We will return to the discwion of the effects of kinetic constrainta on the dynamics of adsorbed chains in the next section. Because of the kinetic constrainta, the predominant eventa in the time period from 10-7 to 10-5 s are the adsorption of free m e n t a that are at the beginning of loops and tails. Pictorially, the adsorption of a long tail can be d&bed as a "zippeming" process: if the firat segment of a tail is number 980, say, the most probable sequence of adsorption eventa within the tail will be 980,981, ..., 999,lOOO. Other event8 may accur during this time period, however, in the other tail or in other loops along the chain.

Macromolecules, Vol. 26, No.5, 1993

Dynamics of PMMA Chains Adsorbed on Al Surfaces 1129

0.015

0.025 (a)

(b)

1

0.020 -

c$

0.015 -

a, 0 C

.-0

5 0.010 E

LL

0.005 -

0 10.l~

I

1 10”

10‘~

1

Time (s)

Time (s) 0.025

0.020

) 5 1 0 ’ 0 v) ”

9a,

0.015

* 0 C

.-0

5 0.010

e

LL

0.005

10-1

Time (s)

10’0

Time (s)

Figure 4. Fractional distribution of individual adsorption-desorption eventa for PMMA chains of 1000 segments adsorbed on an aluminum surface: (a) isotactic chain at 300 K; (b) syndiotactic chain at 300 K (c) isotactic chain at 2000 K; (d) syndiotactic chain at 2000 K. At 300 K,EJkeT = 20; at 2000 K, EdkeT = 3. See the text for the discussion of renormalizing the temperature scale with the activation energy E,.

The final stage of the dynamics at low temperatures, from 10-5 s onward for the isotactic chain at 300 K, occurs through a mechanism different from the one described above. Aa noted above, the intermediate stage of the adsorption of loops proceeds from the two anchoring segments (at the termination of trains) to the interior. As a result, in the final stage of adsorption,the polymer chain configurationwill be comprised of several long trains with a few isolated desorbed (free) segments. These free segmentsare the remnants of loops that have adsorbed all but one of their original members. At low temperature, each loop that is present in the initial confiiation will ultimately leave one of these ‘loop remnants”. When only one free segment of the loop remains, tbl) mechanism of the adsorption changes. Although the adsorption of a segment ie facilitated when one of ita nearest neighbors isalready adsorbed,we find that the adsorptionis hindered when both nearest neighbors are adsorbed. Again, this verifies an assumption of the kinetic Ising models of adsorbed chain dynamics? that steric hindrance and toreional strain retard adsorption rates when both neighbors are adsorbed. For the isotactic PMMA chain, we find the following sequence of events to be most likely for adsorption of the loop remnants 3,3,4,3,3, 3,3,2,3,3,. ...,3,3,3,3,3, where the integers are the labels assigned to the discrete

...,

-

... ...,

..

-

...

states in section 2, and the dots imply that the polymer chain continues in both directions. The presence of the segmentswith only the ester oxygen atom adsorbed (state 2) as an intermediate causes the maximum in the segment population of that state at 10-5 s that can be seen in Figure 3a. The dynamics of adsorbed syndiotactic PMMA at 300 K share several of the qualitative features found for isotactic PMMA. The event distribution for the syndiotactic PMMA, in Figure 4b, shows several distinct time regimes that are very well separated. At times lese than 10-7 s, the events correspond to conformationalrearrangement of segments that were initially adsorbed. Virtually no events occur for 2 decades in time, until the adsorption of loops and tails begins around 10-48. The onset of loop and tail adsorption is seen in the sharp rise of the average train length (FigureSb) and the number of stronglybound segments and the sharp decrease of the number of free segments (Figure 3b). The adsorption of loops and tails for syndiotacticPMMA proceeds in the manner described for isotactic chains. After the zippering process of loop and tail adsorption is nearly complete by 10-2 8, there is another lag in the adsorption events until the loop remnants relax to the strongly adsorbed states. The lag can be seen clearly in the event distribution; it is ale0 evident in the plateaus of the segment populations and

Macromolecules, Vol. !26, No.5,1993

1130 Shaffer and Chakraborty 1 o3

I

(bi

I

t

;

L

t

r 1 10’12

10.8

10’10

10.6

1 0.4

1

I

10.”

10-9

Time (s)

I

I

10-6

1o

I

.~

1

Time (s) lo3

I (4 L

,L

1

I

~~

IO”

10’10

I 0.9

Time (s)

-

10 l 2

10’”

lO”0

10’9

Time (s)

-

Figure 6. Time evolution of the average train length for PMMA chains of loo0 segmenta adsorbed on an aluminum surface: (a) isotactic chain at 300 K; (b) syndiotacticchain at 300 K; (c) isotactic chain at 2000 K (d) syndiotactic chain at 2OOO K. At 300 K, EJkeT 20; at 2000 K, EJkaT 3. See the text for the discussion of renormalizing the temperaturescale with the activation energy E,.

averagetrain length. With the syndiotacticPMMA chain, we find two important paths for the relaxation of the loop remnants:

-

-

...,3,3,4,3,3,... ...,3,3,1,3,3,... ...,3,3,3,3,3,,.. and

-

-

...,3,3,4,3,3,... ...,3,3,2,3,3,... ...,3,3,3,3,3,... The second steps in the paths above are so fast, however, that there is no observable increase in the number of segments occupying states 1 and 2. Hence, there is no maximum in the populations of the intermediate states as seen for the isotactic chain. The dynamicsof adsorbed chains at higher temperatures are very different from the dynamicsat low temperatures. Parte c and d of Figures 3-5 preeent the data for isotactic and syndiotactic PMMA chains adsorbw at 2000 K. Although the high temperature is unrealistic for PMMA chaina, we focus on the QUantitY EJkBT diecueeed above; at 2000 K,EJkBT = 3. Thie value is repreaentative of a more weakly interacting phyoisorbed polymer at room temperature. As expected, at a higher temperature, the adaorption is much faster; it is complete in less than 10“ a. There are no dietinct time regimes, however, as seen by the broad event distributions in parts c and d of Figure 4. The segment populations and the average train length also change gradually; the adsorption of loops and tails

through the zippering process and the subsequent relaxation of loop remnants are no longer the predominant dynamical mechanisms. At the higher temperatures, the effects of the activation barriers to admrption and desorption events are greatly reduced; the kinetic constraints are not severe, and the dynamics are much less cooperative. Thia n+t is expanded in the discussion below. Quantitative Characterization. The qualitative features of the dynamics of PMMA chains adsorbed on aluminum show a degree of cooperativity during the adsorption of loops and tails at low temperatures. The phenomenologicalkinetic Iaing models for adsorbed chain dynamics8 also%howthat dynamical behavior is cooperative and is similar to that observed for glees-forming liquids. Experimental evidence of similarities between adsorbed polymers and structural glassesale0exists.9One major purpose of the present work is to examine the dynamical behavior of a real polymer-solid interface without making assumptions about the nature of the interactions. The goal is to verify the exiebnce of the glaeelike behavior that the phenomenological modele predict8and, more importantly,to examinethe microscopic origins of the cooperative dynamics. In order to characterize the dynamics of adsorbed polymersquantitatively, we have focused on the nonlinear

Macromolecules, Vol. 2f3, No.6, 1993

Dynamics of PMMA Chains Adsorbed on AI Surfaces 1131 1.o

(bl

0.6

1

tr

=I 0.4

t

I

Time (s)

Time (s)

I

0.2

0

10'2

I

I

10'0

10.8

106

Time (s)

Time (s)

Figure 6. Time evolution of the relaxation function defined by eq 25 for PMMA chains of loo0 segments adsorbed on an aluminum surface. The data from the s t o c h t i c simulations (0)are averages over 10oO realizations using independent initial configurations; error bare denote 1 standard deviation of the data. Solid lines are fits to eq 1. The parameters j3 and 7KWW were determined through a weighted nonlinear least-squares fit of the relaxation f~nction:~' (a) isotactic chain at 300 K, @ = 0.412,~ K W W= 2.06 X 1V a; (b) syndiotacticchain at 300 K, @ 0.358, TKWW = 2.71 x lC3S; (c) isotactic chain at 667 K, j3 = 0.562, TKWW = 1.13 x 1O-g B; (d) syndiotactic chain at 667 K, j 3 = 0.425,TKWW = 3.55 X 8.

relaxation function, q(t), for the strongly adsorbed segmenta:

where n(t) is the number of strongly adsorbed segments (thosewith both chemisorbingfunctionalgroups adsorbed) at time t and nY,is the number of strongly adsorbed eegmenta at equhbrium. The relaxation function is normalized to unity at t = 0 and decays to zero if the system evolves to equilibrium. Since we are examining a nonequilibrium praceas, q(t) is not temporally homogeneous; Le., the time origin in eq 25 cannot be shifted arbitrarily as it can be in equilibrium systems. We also calculatethecorresponding nonlinear relaxation time, 4, of the Strongly adsorbed segments:

It is appropriate to define the relaxation time based on the etmngly adsorbedm e n t a becausethe slowest eventa in the adsorption proceea, the relaxation of loop remnants d i e c u d above, involve traneitiona into the strongly adsorbed atate. Thua, 4 is a measure of the longest relaxation time.

The relaxation function q ( t ) and the relaxation time 4 depend on the initial configuration of the polymer chain. Accordingly,its computation requires averagesover many initial configurations and many realizations of the stochastic dynamics. In the results presented below, the initial configuration space was sampled extensively by accumulating averages from lo00 independent initial conditions. For each starting configuration, the total fraction of initially adsorbed segments was chosen randomly from a uniform distribution between 0.05 and 0.50; each of the three adsorbed states was equally likely. The evolution of the relaxation function for isotactic and syndiotactic PMMA at 300 and 667 K ia shown in Figure 6. At 667 K, the normalized activation energy E,,/ ~ B T 9. The behavior of q(t) mirrors the qualitative

-

features of the adsorption that were discussed above. The

different time regimes are evident for both tacticitiea and both temperatures. At first, the initially adsorbed segments relax quickly into the strongly adsorbed configurations; this is responsible for the decrease of q from unity to roughly 0.75. A plateau region of very little activity follows for several decades in time.39 The sharp decrease of q followingthe plateau region correspondsto the second stage, adsorption of loops and tails. By far the most adsorption eventa occur in this region where q decreases

1132 Shaffer and Chakraborty

from 0.75 to 0.25. The final stage, in which the loop remnants relax, is especially apparent in Figure 6b because the time regions are very well separated. (See the event distributions of Figure 4b.) An important result of the quantitative characterization is that the decay of the relaxation function is not purely exponential in time. Except for the very fmt and last periods of adsorption, the relaxation function can be described by the stretched-exponential or KohlrauschWilliams-Watts (KWW) function given in eq 1. (Note that f K W W is not equivalent to the nonlinear relaxation time 4,)The KWW function fits the relaxation function well only in the time range that corresponds to the adsorption of loops and tails (the overwhelming majority of events). In the time regions where the fit is not satisfactory, the physical mechanisms that lead to adsorption are different. In the initial time period the dominanteventsare local conformationalrearrangements, and in the final time period the dominant events are the adsorption of loop remnants. The nonexponential relaxation that we observe over a wide range of temperatures is a direct consequence of the kinetic constraintsthat are operativeduring the adsorption of the loops and tails. The stretching parameter /? in eq 1 is a rough indicator of the severity of the kinetic constraints. (See the discussion below.) Therefore] it is instructive to examinehow the stretching parameter varies as a function of temperature. The variation of /? with inverse temperature is shown in Figure 7 for both isotactic and syndiotactic chains. The values of B nearly coincide for the two tacticities at the highest and lowest temperatures but differ at intermediate temperatures. At the highest temperature] where EdkBT = 1,the values of /? are rather large, around 0.75. At infinite temperature, one expects that /? would asymptotically approach unity (Debye relaxation). As the temperature is lowered, the values of /? decrease until they finally plateau near 0.35 for temperatures below 300 K (EJkBT > 20). The results presented above show that the dynamical behavior of PMMA chains adsorbed on aluminum is qualitatively similar to that observedfor structural glasses. Thus, we expect that the physical mechanismsresponsible for the dynamical features of strongly adsorbed polymers are similar to those present in structural glasses. It has been shown that dynamical constraints give rise to nonexponentialrelaxati0n;40-~2these constraints generally result from rugged potential energy or free energy surf a c e ~ . ' PMMA ~ ~ ~ ~ chains adsorbed on aluminum and structural glasses share both of these features: rugged energysurfacesand strong kinetic Constraints. In previous kinetic Ising models of bulk polymers,17adsorbed polymers? and structural glasses,44dynamical constraintswere introduced explicitlywithout making a detailed connection to their microscopic origin. In other words, the kinetic constraints that are operative were not derived explicitly from the microscopicHamiltonian. Nevertheless,the fact that these modela reproduce the qualitative features found in real polymers and glass-forming liquids enforces the idea that dynamical constraints are responsible for the g h y dynamice. For the stronglyadsorbedPMMA chains, however, the microscopic origins of the dynamical constraints are much clearer. For PMMA adsorption at low temperatures,strongkinetic conetraintaare in effect during the adsorption of loops and tails. As discussed above, the adsorption rate of a PMMA segment is greatly affected by ita nearest neighbors and entropic penalties imposed by the surface upon adsorption of segments further along the chain. As a consequence,the dynamicsof the adsorbed

MaCrOmOkCUkS, 1.0

1

0.2

cII i-

*

i I

i

t 0' 0

Vol. 26, No.5,1993

I

'

" 2

I

'

" 8

1

6

4

'

J 10

1OOOm (K.') 1.o

08

(W

!

ao6b I*

T

04

02

T

T

T

c

01 0

'

I

2

'

4

'

1

6

'

'

8

'

1

10

10001T(K")

Figure 7. K W W stretching parameter 6 (0) versus inverse temperature. Error b m indicate 1 standard deviation of the fitted parameter:" (a) isotactic chain;(b)ayndiotacticchain. The solid line in (b) is the function 6 = 0.36 + 0.66 exp(-lW/T).

chains are highly cooperative. For instance, before a segment near the end of a tail or the middle of a loop has a significant chance to adsorb at low temperatures, all of the segments preceding it back to the anchoring segment must adsorb. The role of the dynamical constraints in strongly adsorbed polymers can also be seen by considering their effect on the topology of the configuration space of the adsorbed pdymer.11t41*42The polymer dynamics can be described by the time evolution of the confiiation point Q among the 4N discrete pointa in configuration space. The transition rates between configuration space pointa are governed by the structure of the free energy surface, i.e., the adeorption energetica and the conformationaland configurational changes that accompany adeorptiondesorption eventa. The strong and orientation dependent segment-surfaceenergeticslead to kinetic bottlenecksfor conformationaland confiiati& transitions that must occur for relaxation to the equilibrium adeorbed state. For the PMMA-aluminum system at low temperatures, many transitions between pointa in the confiiation space are eesentiallyforbidden. In effect,the kinetic conetaainta reduce the availablepaths between pointa in d i a t i o n space to a sparsely connectad network. hlaxation to equilibrium then must follow a tortuous route through the configuration space. Thia ia the origin of the cooperative dynamics that we observe. Computer simulations

MaCrOmOkCUle8,

Vol. 26, No.5, 1993

Dynamics of PMMA Chains Adsorbed on Al Surfaces 1133

of several models with restricted paths through configuration space do show the effects of kinetic constraints through nonexponentialr e l a x a t i ~ n In . ~the ~ PMMA~~~~ aluminum system, the kinetic bottlenecka which arise naturally from the free energy surface and which lead to cooperative dynamical behavior also lead to nonexponential relaxation. It is interesting to consider what physical significance to attach to the stretching parameter 6 in the KWW function. Intuitively, one expects that the value of 6 is a rough indicator of the severity of the kinetic constraints, or the degree to which they are operative in a given situation. The decrease of 6 with temperature can be interpreted in thisway. At high temperatures, the kinetic constraints involved in loop and tail adsorption are not as important because the features of the potential energy 0 2 4 6 0 10 surface that give rise to the constrainta become small in lOOO/T (K-') comparison to the thermal energy, kBT. As the temperature is lowered, energy barriers for many adsorption 106, pathways become significantly larger than kBT. These pathways are then effectively eliminatedas possible routes to relaxation, and more cooperativemotions are required. The plateau of 6 below 300 K indicates that the kinetic constrainta are fully operative at 300 K (EdkBT 20); loweringthe temperature has no further observable effect on the severity of the constraints. Similar behavior has been seen in a simplified model of glassy relaxation.&The diffusive motion of a random-walker on a hypercube with randomly deleted vertices has been described with the KWW function; the value of 6 decreases from unity to roughly l/3 as the network of paths becomes increasingly sparse.46 The temperature dependenceof the nonlinear relaxation times is also important for relating the dynamics of adsorbed polymers to dynamics in glass-formingliquids. 1 2 3 4 5 0 At each temperature, a relaxation time was calculatedfrom 1OOOR (K-') eq 26 for each of the lo00independent realizations of the Figure 8. Arrhenius plot of the nonlinear relaxation times, PI, adsorption process. The upper limit of integration was defined by eq 26. The data from the stochastic simulations ( 0 ) the time required for the relaxation function to decay to are averagesover loo0 realizations; error bars denote 1 standard 0,001;this corresponds to the relaxation of all but one deviation of the data. Solid lines are fita to the Arrheniua form, segment in the chains of lo00 segments. The variations eq 3, wing the relaxation times at temperatures 300 K and below: (a) isotactic chain, E h = 42 kJ mol-'; (b) syndiotactic of the nonlinear relaxation times with temperature are chain, E h = 67 kJ mol-'. shown in the Arrhenius plots of Figure 8. For a simple relaxation process with a single significant activation s, and TVF= 75 K. While we know that the true behavior energyE h , therelaxation time should follow the Arrhenius below 200 K is Arrhenius, the long relaxation times might expression of eq 3. For the syndiotactic chain, Arrhenius behavior is seen at all temperatures below lo00K ( E ~ ~ B Tnot be accessible to experiments or detailed computer simulations. Without the knowledge of the relaxation =6 ) . For the isotactic chain,the temperature dependence times at lower temperatures, the fact that the data are is non-Arrhenius at high and intermediate temperatures consistent with the Vogel-Fulcher equation might be but reverts to Arrhenius behavior at temperatures below interpreted as foreshadowing a real divergence at TW = 200 K 30). From the slopes of the lines in 75 K. Figure 8, the apparent activation energy E k is estimated The temperature dependence of the relaxation times to be 42 kJ mol-' for isotactic PMMA and 67 kJ mol-' for for isotactic PMMA in the non-Arrheniusregion indicates syndiotactic PMMA. These values are within the range that the apparent activation energy for relaxation, of the "baeeline" activation energy, E,, of 50 kJ mol-' that has been used to normalize the temperature scale. d(ln 7') E*(T)= Although the behavior at low temperatures is Arrhenius, 41/79 it should be emphasized that the relaxation times at the lowesttemperatumare beginning to surpassexperimental is increasing as the temperature is lowered from 2000 to time scales. For the isotactic chain at 100K,for example, 300 K (as EJkBT increases from 3 to 20). The increase 4 is approximately 33 yeara. As a result, experimenta in the apparent activation energy has two separate causa. that cannot access these time scales will conclude that the Firat,the local transition probabilities,RL(Q&+dQi-&+l), relaxation times have diverged, corresponding to a dydecrease with decreasing temperature. This is a general namical falling out of equilibrium and the strongly nonfeature of systems with rugged potential energy surfacArrhenius behavior suggested by the Vogel-Fulcher form es.12J3 According to eq 8, each transition path from one of eq 2. In fact, the relaxation times for the isotactic chain macrostate to another is weighted by the probability of above 200 K can be fit well by the VogekFulcher form the microstate originating the transition being occupied. with the parameters EW = 25 kJ mol-', w = 7.3 X 10-12 As the temperature is lowered, fewer microstates will be

-

I"

-

1134 Shaffer and Chakraborty

Macromolecules, Vol. 26, No. 6, 1995

impenetrable surface. In other word% the partially occupied with a significant probability, so the number of adeorbedPMMAdrainep~~deto~logicalo~~l~that paths contributing to the overall transition probability impedethe desorptionof PS chaina As a result, the kinetic will decrease. This reflects an entropic restriction at the constraintsthemaelveeevolvein time asthe PMMA chaina level of individual polymer segments (unrelated to the relax toward equilibrium; the theoretical treatment of entropic constraints imposed by the surface) that grows competitive adsorption would necessarily be more commore severe as the temperature is lowered. Furthermore, plicated than the analysis presented here. one expects the low energy microstates (which will be occupied preferentially as the temperature is lowered) to 6. Conclusions have higher activation energies for significant conformationaltransitions (thosethat changetheadsorption state). We have presented a stochastic model for the dynamics So not only are the number of transition paths reduced of PMMA chains adsorbed on an aluminum surface. The as the temperature is lowered but also the paths with the model can investigate the length scales and time scales highest activation energies are likely to be the ones of relevant to polymer adsorption by employing a coarseimportance. Second, as discussed above, the topology of grained representation of the polymer chain (in terms of the polymer c o n f i a t i o n spacechangeswith temperature. discrete adsorption states) and stochastic dynamics. At higher temperatures, relaxation can proceed by many Previous kinetic Ising models of polymer dynami~s811~ have routes in addition to the zippering of loops and tails also used discrete-state representations of polymer condescribed above. As the temperature is lowered, however, figurations. The model presented here differs from the the kinetic constraints become active, and many paths in kinetic Ising models in two respects. In the present model, configuration space become kinetically blocked. This polymer segmentsare classified into four discrete s t a h , rarefaction of the accessible paths in configuration space rather than the two-state classification used in Ising is equivalent to a gradual increase in entropic barriers to models. The crucial distinction of the present model, relaxation on the scale of the configuration space of the however, is that the dynamical rules are not phenomeentire polymer chain. As the paths become blocked, more nological but are derived directlyfrom the potential energy cooperativity is required for relaxation to equilibrium. As surface. In the kinetic Ising model of adsorbed polymer in other systems with strong kinetic ~ o n s t r a i n t s , 4 ~ * dynamics? ~ ~ ~ ~ ~ ~ kinetic constraints operating between neighthe requirement for cooperativedynamica over long length boring segments were postulated based on physical inscales leads to non-Arrhenius behavior. duction. In particular, the authors assumed that the In contrast to the non-Anhenius temperature depenadsorption rate of a particular segment along the chain dence for the isotactic chain at high temperatures, the would be enhanced if one of its nearest neighbors was region of Arrhenius dependence indicates that the charalready adsorbed. Here, adsorption rate constante calacter of the final stage of relaxation differs from that of culated from the potential energy surfaceyielded the same the region in which stretched-exponential relaxation is qualitative features of the kinetic constraints that were observed. The Arrhenius behavior at low temperatures used in ref 8. This finding verifies that the qualitative reflects the nature of the final stage of the relaxation, the behavior of the kinetic king model illustrates general adsorption of loop remnants. Since 7"' is a measure of the features of strongly interacting polymer-eolid interfaces. longest relaxation time of the adsorbed chains, at low By incorporating the details of the PMMA-aluminum temperatures its behavior is dominated by the slowest potential energy surface, the stochastic model has been dynamicalprocess, the adsorptionof loop remnants. Since able to investigate the effects of tacticity on the adsorbed the loop remnants are spatially separated along the chain dynamics. There are many qualitative similarities polymer chain, they relax independently of each other; between the dynamicaof isotactic and syndiotacticPMMA cooperative motions over large length scales, which often chaina adsorbed on aluminum;nevertheleas,the stochastic lead to non-Arrhenius behavior, are not required in the model predicts severalquantitative ef€& of tacticity. At final stage of relaxation. Furthermore, at low enough temperatures below 500 K,relaxation times for syndiotemperatures, the single dynamicalpathway (witha single tactic chains are several orders of magnitude greater than activation energy) that corresponds to relaxation of loop those for isotactic chains and the apparent activation remnants is the rate-limitingstep in the relaxation process; energy for relaxation of a syndiotactic chain is roughly 1.5 the activation energy for that pathway is the energy Ehr times that for an isotactic chain. that appears in eq 3. The dynamics of PMMA adsorbing on aluminum are Many of the phenomenological features of the dynamics qualitatively similar to the dynamics of glass-forming of strongly adsorbed polymers have been demonstrated liquids. Both system exhibit nonexponential relaxation through competitive adsorption experiments. Johnson and non-Anheniustemperature dependencesof relaxation and Granickshave studied the kinetics of the displacement times. The common features are rugged energy surfaces of polystyrene (PSI on oxidized silicon by PMMA. The and strong kinetic constraints. While the exact microdesorption kinetics were described well by a KWW scopic basis of the kinetic constraints in glass-forming function,eq 1,in which both @ and TKWW were temperature liquids remains unclear, the present stochastic model has dependent. The value of 6 increased from 0.5 at 25 O C to clarified the origins of the kinetic constraints for PMMA 0.6 at 50 "C. The temperature dependence of TKWW was chains adsorbed on aluminum. For strongly adwrbed clearly non-Arrhenius; the time required for desorption polymers, the kinetic constraints have several phyleical exceeded experimental time scales below 25 "C. The origins: (1)nearestneighbor interactions; (2) strong and experimental situation of competitive adsorption is somespecific segment-surface interactions; (3) entropic conwhat different than the caee of adsorption from dilute straints introduced by an impenetrable surface; (4) tosolution treated here. We expect, however, that the pological barriers due to other adsorbed chaine,and (5) underlying idea that glassy dynamics multa from strong competitionfor surfacesites. The fmt twoeffecte depend kinetic constraints is still applicable. As discussed by on the detaila of the interaction potential, and thus the Johnson and Granick, for PS chains to desorb from the specific chemical constitution of the polymer segments and the substrate; the other effects are generic to all surface, they must diffuse through a network of 'cages" formed by the strongly adsorbed PMMA chains and the polymer systems. Topologicalconstraints are responsible

Macromolecules, Vol. !26, No.5, 1993

Dynamics of PMMA Chains Adsorbed on Al Surfaces 1136

for the "caging" effect described by Johnson and Granickg for the glassy dynamics of PS desorption from a surface in the presence of partially adsorbed PMMA. Similar caging effects should also affect the dynamics of PMMA adsorption from concentrated solutions onto initially bare surfacesbecaw segmenta in loops and tailscannot diffuse freely to the surface in the presence of other partially adsorbed polymers. Competition for surface sites may also provide significant kinetic constrainta. Once a large fraction of the surface is covered with adsorbedsegmenta, a segment still in a short loop or tail may have to wait for adsorbed segmenta to diffuse in the surface layer and free a surface site. The model presented here explicitly accounta for only the fmt three effects. Although the last two effects are most important for concentrated solutions and melta, they can also be important for an isolated chain, which may see other parte of itaelf as obstacles to adsorption and competitors for surface sites. However, the additional complexities should only serve to make the effecte that we have observed more pronounced. Work currently underway is aimedat investigating the dynamical consequences of topological constrainta and competition for surface sites. Acknowledgment. This material is based on work supported under a National Science FoundationGraduate Fellowship awarded to J.S.S. A.K.C. thanks the Shell Foundation, the Dexter Corp., and the U.S. DOE Office of Basic Energy Sciences for fiancial support. Computational resources were provided by the University of California Program in Supercomputing at UCLA. Discussions with W. H. Miller and E. M. Sevick regarding the search for potential energy extrema are gratefully acknowledged. Appendix Here we show that the effective rate constants defined by eq 8 satisfy the condition of detailed balance for the macrostate populations. The statement of detailed balance is

(28) where and are the true equilibrium populations of macrostates A and B, respectively. The equilibrium population of macrostate A is given simply by

e

e

zA/z

(29) where ZA is the local configurational partition function defied by eq 7 and Z (without a subscript) denotes the configurational partition function for the entire configuration space:

z = JeXp[-V(x)/kBT]

dx

(30)

The individual rate constanta for transitions between microstates satisfy detailed balance among themselves: PPki-j = P?kj-i (31) where p p and pi" are the true equilibrium probabilities for microstates t and j: p p = z'JeXp[-v(%)/kBT] dn (32) As in eq 6,the integration in the preceding equation is limited to the volume of configuration space enclosing microstate i. By comparing eq 32 with eq 6,the following

identities are clear:

(33) The ratio of the macrostate rate constanta can be written as follows:

where the summations are over all transition between macrostates, as in eq 8. Now substitute the identities from eq 33,yielding

Because the individual rate constants ki-j satisfy detailed balance, the following equation holds:

(36) and the ratio of the summations in eq 35 is unity. That leaves

(37) in accord with the condition of detailed balance, eq 28. References and Notes (1) Scheutjens, J. M. H. M.; Fleer, G. J. J. Phys. Chem. 1979,83, 1619; 1980,84, 178. ( 2 ) Eisenriegler, E.; Kremer, K.; Binder, K. J. Chem. Phys. 1982, 77,6296. (3) de Gennes, P.-G. Ado. Colloid Interface Sci. 1987,27, 189. (4) Takahashi, A.; Kawaguchi, M. Ado. Polym. Sci. 1982,46, 1. (5) Cohen-Stuart, M.; Cosgrove, T.; Vincent, B. Ado. Colloid Interface Sci. 1986,24,143. (6) Shaffer, J. S.;Chakraborty, A. K.; Tirrell, M.; Davis, H. T.; Martins, J. L. J. Chem. Phys. 1991,95,8616. (7) Chakraborty,A. K.;S M e r ,J. S.;Adriani,P. M.Macromolecules 1991,24,5226. (8) Chakraborty, A. K.; Adriani, P. M. Macromolecules 1992,26, 2470. Adriani, P. M.; Chakraborty, A. K. J. Chem. Phys., - . in press. Johnson, H. E.; Granick, S. Science 1992,255,966. Brawer,S.Reluxation in Viscous Liquids and Glasses;American Ceramic Society: Columbus, OH, 1985. Fredrickson, G. H. Annu. Reo. Phys. Chem. 1988,39,149. Hoffmann, K. H.; Sibani, P. Phys. Rev. A 1988,38,4261. Frauenfelder, H.; Sligar, S. G.; Wolynes, P. G. Science 1991, 254, 1598. Theglassy interfaciallayerdiscussedhere shouldnot be confused with a glassy layer near an interface due to increased density, as proposed by: Kremer, K. J. Phys. (Paris) 1986,47, 1269. The effecte alluded to here are purely kinetic, introduced via strong and specific interactionswith the surface, and have no relationshipto the glass transition of the bulk polymer. Chakraborty, A. K.; Davis, H. T.; Tirrell, M. J. Polym. Sci., Polym. Chem. Ed. 1990,28,3185. As general references for kinetic Ising models, see the following: Glauber, R. J. J. Math. Phys. 1963,4,294. Kawasaki, K. Inphase Tmnaitiona and Critical Phenomena; Domb, C., Green, M. S.,Me.; Academic Press: London, 1972; Vol. 2. Other kinetic Ising models for polymer dynamics can be found in the following: Orwoll,R. A.; Stockmayer, W. H. Ado. Chem. Phys. 1969,15, 305. Skinner, J. L. J. Chem. Phys. 1983, 79, 1955. Isbister,D. J.; McQuarrie,D. A. J. Chem. Phys. 1974,60, 1937. Budimir,J.;Skinner, J. L. J. Chem. Phys. 1986,82,5232. Frantz, P.; Granick, S.Phys. Reo. Lett. 1991,66,899. van Kampen, N. G. Stochastic Processes in Physics and Chemistry; North-Holland: Amsterdam, 1981. Stillinger, F. H.; Weber, T. A. Phys. Reo. A 1982,25,978. June, R.L.; Bell, A. T.;Theodorou, D. N. J. Phys. Chem. 1991, 95, 8868.

Hfinggi,P.;Talkner, P.;Borkovec,M. Reo. Mod.Phys. 1990,62, 251.

1136 Shaffer and Chakrahrty Palmer, R. G. Adv. Phys. 1982,31,669. Chandrasekhar, s. Rev. Mod.phy8. 194,15, 1. Hoeve, C. A. J. J. Chem. phys. 1966,43,3007. Feller, W. An Zntroduction to Probability Theory and Its Applications; wiby New York, 1968. Tsikoyiannis, J. G.; Wei, J. Chem. Eng. Sci. 1991, 46, 233. Hoel, P. G.; Port, S. C.; Stone,C. J. Zntroduction to Stochastic h e 8 s e 8 ; Houghton M i Boston, 1972. Vacatello, M.;Flory,P. J. Macromo~ecu~e8 1986,19,405. Nicholeon, D.; Paraonage, N. G. Computer Simulation and the Statieticali%fechanicsofddsorption;AcademicPress: London, 1982. de Heer, W. A.; Milani, P.; ChAtelain, A. Phys. Rev. Lett. 1989, 63,2834. Cerjan, C. J.; Miller,W.H. J. Chem. Phys. 1981, 75,2800. Chang, G.; Guida, W. C.; Still, W. C. J. Am. Chem. SOC.1989, 111,4379. Banerjee, A.; Adams, N.; Simon, J.; Shepard, R. J.Phys. Chem. 1988,89,52. Baker, J. J. Comput. Chem. 1986, 7,385. Fhanberg, R. 0.;Berne, B. J.; Chandler, D. Chem. Phys. Lett. 1980,75,162. Montgomery,J. A., Jr.; Holmgren,S. L.; Chandler, D. J. Chem. Phys. 1980, 73,3688. Chandler, D.J. Chem. phy8. 1978,68,2959. Voter, A,; Doll, J. D. J. Chem. Phys. 1986,82,80. Even after performing lo00 simulations it is difficult to accumulate good statistics in the region of time between the

Macromokculee, Vol. 26, No.5, 1893 relaxation of initially adsorbed wgmenta and the dmrptionof loops and tails. The poor s h t i ~ t i c are a responsible for the noisy appearance of the relaxation fuactio~in thin region. (40)Palmer, R. G.; Stein, D. L.;Abrahams, E.;Andemon, P. W. PhyS. Rev. Lett. 19&(,53, 958. (41) Palmer, R.G. In Heidelberg Colloquium on G h 8 y Dyn5mics; van Hemmen, J. L., Morgenstem, I., Ede.; Springer-Verb Berlin, 1987. (42) Palmer, R. G.; Stein, D. L. In Lectures in the Sciences of Complexity; Stein, D. L., Ed.; Addison-Wesley: Redwood City, CA, 1989. (43) Note that a rugged free energy surface cm be obtained from a structureleapotential energy surfax through the introduction of dynamid constraints like forbidden pathways. The dynamical c o ~ t r t h t eintroduce entropic barriers on the free energy surface. See ref 41. (44)Fredrickson, G. H.; Andexsen, H. C. Phys. Rev. Lett. 1984,53, 1244, J. Chem. PhyS.l9&83,5822. (45) Campbell, I. A.; Fleaaelles, J.-U; Jullien, R.; Botet, R.J. PhYS. c 1987,20, L17; Phys. Rev. B 1988,37,3825. (46) Fredrickeon, G. H.;Brawer, S. A. J. Chem. Phye. 1986,84,3351. (47) Press, W. H.; Flnnnery,B. P.; Teukolsky,S. A.; Vettorling, W. T.Numerical Recipes;CambridgeUniversity P m : Cambridge, U.K., 1986.